NetApp SnapMirror - Riverbed · PDF filePERFORMANCE BRIEF NetApp SnapMirror Riverbed...
4
PERFORMANCE BRIEF NetApp SnapMirror Riverbed Steelhead ® Appliances Accelerate NetApp SnapMirror ® To provide maximum protection and ease of management, many enterprises choose to perform SnapMirror operations across the wide-area network. However, WAN links are often costly. Furthermore, the performance of SnapMirror is often severely degraded by the limited bandwidth and high network latency typical of a WAN. Due to the large amount of data to be transferred, a task such as mirror initialization may take days to complete over a WAN. Applications that use filers for storage typi- cally touch large numbers of storage blocks as files are added, deleted, and modified during a typical workday. This results in the marking of many blocks on the filer for mirroring, which means that incremental updates to remote mirrors may take hours to complete, and often cannot be completed during the available backup window. Steelhead-Enhanced SnapMirror The network traffic associated with both mirror initialization and updates contains a large amount of compressible, repetitive data, even accounting for SnapMirror’s pow- erful differencing algorithms. The Steelhead’s SDR, VWE, and compression features can normally remove a huge amount of additional repetitive data from the WAN during these operations, which dramatically reduces the time to complete mirror initialization, updates, and recovery. Typical backup times can often be reduced by 90% (see graphs). Performance Improvements Lab tests show that Steelhead appliances significantly improve the performance of SnapMirror over the wide area network. Cold mirror backup is approximately 30% faster, while warm mirror backup or recovery is a dramatic 45 times faster. Incremen- tal backup is 8 times faster. In production environments, incremental updates are the most critical, as they are representative of the bulk of information traversing the network. Bandwidth utilization is reduced by up to 89% during incremental updates, implying that Steelhead appliances eliminate approximately 9 out of every 10 bytes that previously crossed the WAN. TEST RESULTS Steelhead appliances improve the per- » formance of SnapMirror up to 45x Incremental updates are 8x faster » Bandwidth utilization is reduced by » up to 89% TESTING SCENARIO SnapMirror operations were performed on a 5GB volume across a WAN environ- ment that consisted of a T1 connection and 100 milliseconds of delay between locations. This scenario is representative of a standard New York to Los Angeles WAN connection. The tests consisted of three parts, reflecting common use cases: Initializing a remote mirror across • the network Restoring the mirror back across the • network Backing up the remote mirror after • a 10% change to the data in the volume A “Cold Run” is defined as a data transfer that has never been seen by the Steelhead appliance before (a com- pletely new file). A “Warm Run” is defined as a data transfer in which the Steelhead appli- ance has seen most or all of the data before (a mirror recovery or an incre- mental update). Full Datastore Initialization or Recovery (5GB Directory Tree) – Time to Complete (in minutes) More than 45x faster 0 100 200 300 400 500 600 Without Steelhead Cold Run Warm Run
Transcript of NetApp SnapMirror - Riverbed · PDF filePERFORMANCE BRIEF NetApp SnapMirror Riverbed...
PERFORMANCE BRIEF
NetApp SnapMirrorRiverbed Steelhead® Appliances Accelerate NetApp SnapMirror®
To provide maximum protection and ease of management, many enterprises choose to perform SnapMirror operations across the wide-area network. However, WAN links are often costly. Furthermore, the performance of SnapMirror is often severely degraded by the limited bandwidth and high network latency typical of a WAN.
Due to the large amount of data to be transferred, a task such as mirror initialization may take days to complete over a WAN. Applications that use filers for storage typi-cally touch large numbers of storage blocks as files are added, deleted, and modified during a typical workday. This results in the marking of many blocks on the filer for mirroring, which means that incremental updates to remote mirrors may take hours to complete, and often cannot be completed during the available backup window.
Steelhead-Enhanced SnapMirror
The network traffic associated with both mirror initialization and updates contains a large amount of compressible, repetitive data, even accounting for SnapMirror’s pow-erful differencing algorithms. The Steelhead’s SDR, VWE, and compression features can normally remove a huge amount of additional repetitive data from the WAN during these operations, which dramatically reduces the time to complete mirror initialization, updates, and recovery. Typical backup times can often be reduced by 90% (see graphs).
Performance Improvements
Lab tests show that Steelhead appliances significantly improve the performance of SnapMirror over the wide area network. Cold mirror backup is approximately 30% faster, while warm mirror backup or recovery is a dramatic 45 times faster. Incremen-tal backup is 8 times faster. In production environments, incremental updates are the most critical, as they are representative of the bulk of information traversing the network. Bandwidth utilization is reduced by up to 89% during incremental updates, implying that Steelhead appliances eliminate approximately 9 out of every 10 bytes that previously crossed the WAN.
TeST ReSulTS
Steelhead appliances improve the per-»»formance of SnapMirror up to 45x
Incremental updates are 8x faster»»Bandwidth utilization is reduced by »»up to 89%
TeSTiNg SceNARio
SnapMirror operations were performed on a 5GB volume across a WAN environ-ment that consisted of a T1 connection and 100 milliseconds of delay between locations. This scenario is representative of a standard New York to Los Angeles WAN connection. The tests consisted of three parts, reflecting common use cases:
Initializing a remote mirror across •the network
Restoring the mirror back across the •network
Backing up the remote mirror after •a 10% change to the data in the volume
A “Cold Run” is defined as a data transfer that has never been seen by the Steelhead appliance before (a com-pletely new file).
A “Warm Run” is defined as a data transfer in which the Steelhead appli-ance has seen most or all of the data before (a mirror recovery or an incre-mental update).
Full Datastore Initialization or Recovery (5GB Directory Tree) – Time to Complete (in minutes)
More than 45x faster
0 100 200 300 400 500 600
WithoutSteelhead
Cold Run
Warm Run
DeployMeNT BeNeFiTS
Deploying Riverbed in conjunction with SnapMirror provides multiple benefits, including:
More efficient mirror operations •including initialization, update, and recovery from disasters. Signifi-cantly reduce backup windows and restoration time for even the largest volumes.
More frequent updates.• Given the ability to efficiently mirror incremen-tal data, companies can mirror more frequently.
Disaster recovery across low band-•width links. Steelhead efficiency allows companies to mirror data from locations that were previously infeasible.
PERFORMANCE BRIEF: NetApp SnapMirror
These results are based on the testing scenario presented in this paper. Your results may vary based on the conditions of your own network and the specifics of your own use cases.
Incremental Update (10% change to 5GB Directory Tree) – Time to Complete (in minutes)
Bandwidth Utilization – Incremental Update (in megabytes)
More than 8x faster
0 10 20 30 40 50 60
WithoutSteelhead
Cold Run
Warm Run
n/a
89% less WAN traffic
0 100 200 300 400 500 600 700
WithoutSteelhead
Cold Run
Warm Run
n/a
Steelhead Appliance FeaturesSteelhead appliances leverage a combination of patented data reduction, TCP optimization, and application-level throughput optimizations, as well as remote office file and management functionality, to provide a comprehensive solution for WAN optimization that scales across a broad range of applications and network topologies.
Scalable Data Referencing (SDR) – Riverbed’s SDR algorithms work across all TCP applications including Microsoft Office, Lotus Notes, CAD, ERP, NFS, FTP, and HTTP, to ensure the same data is never sent more than once over the WAN. SDR reduces band-width consumption for many applications dramatically, typically by 60% to 95%, and sometimes more.
Transparent Pre-Population – Appliance data stores can be automatically and trans-parently pre-populated with new file system data or email data to accelerate the initial access to this data by the client.
Application-Specific Optimizations – Steelhead appliances minimize the impact of WAN latencies on applications. By minimizing round trips and payloads generated at the application layer, Riverbed provides additional order-of-magnitude throughput increases to applications including Windows file sharing (CIFS), Exchange (MAPI), Web (HTTP), Database (MS-SQL), FTP, and backup and replication.
Virtual Window Expansion (VWE) – VWE enables applications to overcome TCP win-dowing limitations to dramatically increase the amount of data that can be sent in a single round trip.
High-Speed TCP (HS-TCP) – For high latency, high bandwidth links, unaided TCP often fails to fill the link leaving much of the WAN bandwidth unusable. HS-TCP is available on the Steelhead 5010, and supports up to 750 Mbps per connection for blazing fast data replication and backup.
Proxy File Service (PFS) – PFS enables remote office workers local access to files even when the WAN link to the office goes down. PFS also enables remote office file shares to be replicated automatically to the data center, ensuring reliable backup.
Steelhead Appliances Accelerate a Broad Range of Applications
© 2008 Riverbed Technology. All rights reserved. Portions of Riverbed’s products are protected under Riverbed patents, as well as patents pending. Riverbed Technology, Riverbed, RiOS, Steelhead, Interceptor, and the Riverbed logo are trademarks or registered trademarks of Riverbed Technology. PB-SM031406
25x 100x or more ->0 5x 10x 15x 20x
ERP - Oracle 11i, SAP NetWeaver
ERP - Fat Client
Database - Oracle SQL
Database - MS-SQL
Software Distribution - SMS
Backup & Replication
FTP
Web - HTTP, HTTPS, Intranet
Mail - Notes
SharePoint
Mail - Exchange (MAPI)
File Sharing - Unix (NFS)
File Sharing - Windows (CIFS)
120
10
5
5
20
44
50
Document Management
50
60
Peak PerformanceMost Likely range
50
70
200
55
45
Typical Deployment Architecture
REGIONAL OFFICE OR DISASTER RECOVERY SITE
Steelhead
WAN
Steelhead
DATA CENTER
NetApp Filer NetApp Filer
2005, 2006, 2007, 2008, 2009
About RiverbedRiverbed Technology is the IT performance company. The Riverbed family of wide area network (WAN) optimization solutions liberates businesses from common IT constraints by increasing application performance, enabling consolidation, and providing enterprise-wide network and application visibility – all while eliminating the need to increase bandwidth, storage or servers. Thousands of companies with distributed operations use Riverbed to make their IT infrastructure faster, less expensive and more responsive. Additional information about Riverbed (NASDAQ: RVBD) is available at www.riverbed.com.
Riverbed Technology199 Fremont StreetSan Francisco, CA 94105Tel: +1 415 247 8800Fax: +1 415 247 8801www.riverbed.com
Riverbed Technology Pte. Ltd.391A Orchard Road #22-06/10Ngee Ann City Tower ASingapore 238873Tel: +65 6508-7400
Riverbed Technology Ltd. Farley Hall, London RoadBinfieldBracknellBerks RG42 4EUTel: +44 (0) 1344 401900
Riverbed Technology K.K.Shiba-Koen Plaza Building 9F3-6-9, Shiba, Minato-kuTokyo, Japan 105-0014Tel: +81 3 5419 1990
© 2010 Riverbed Technology, Inc. All rights reserved. Riverbed and any Riverbed product or service name or logo used herein are trademarks of Riverbed Technology, Inc. All other trademarks used herein belong to their respective owners.
PB-SM022809
